Studying Continuous Symmetry Breaking us-

ing Energy Level Spectroscopy

A. Wietek, M. Schuler, and A. M. Lauchli

Institut fur Theoretische Physik
Universitat Innsbruck
arXiv:1704.08622v1 [cond-mat.str-el] 27 Apr 2017

Lecture Notes of the Autumn School on Correlated Electrons 2016, Quantum

Materials: Experiments and Theory, Forschungszentrum Julich (2016)

Abstract

Tower of States analysis is a powerful tool for investigating phase transitions in condensed
matter systems. Spontaneous symmetry breaking implies a specific structure of the energy
eigenvalues and their corresponding quantum numbers on finite systems. In these lecture
notes we explain the group representation theory used to derive the spectral structure for
several scenarios of symmetry breaking. We give numerous examples to compute quantum
numbers of the degenerate groundstates, including translational symmetry breaking or spin
rotational symmetry breaking in Heisenberg antiferromagnets. These results are then com-
pared to actual numerical data from Exact Diagonalization.

Contents
1 Introduction 2

2 Tower of states 3
2.1 Toy model: the Lieb-Mattis model . . . . . . . . . . . . . . . . . . . . . . . . 3

3 Symmetry analysis 6
3.1 Representation theory for space groups . . . . . . . . . . . . . . . . . . . . . . 8
3.2 Predicting irreducible representations in spontaneous symmetry breaking . . . 9

4 Examples 10
4.1 Discrete symmetry breaking . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4.2 Continuous symmetry breaking . . . . . . . . . . . . . . . . . . . . . . . . . . 14

5 Outlook 22
 1 INTRODUCTION

1 Introduction

Spontaneous symmetry breaking is amongst the most important and fundamental concepts in
condensed matter physics. The fact that a ground- or thermal state of a system does not obey its
full symmetry explains most of the well-known phase transitions in solid state physics like crys-
tallization of a fluid, superfluidity, magnetism, superconductivity and many more. A standard
concept for investigating spontaneous symmetry breaking is the notion of an order parameter. In
the thermodynamic limit it is non-zero in the symmetry-broken phase and zero in the disordered
phase.

Another concept to detect spontaneous symmetry breaking less widely known but equally pow-
erful is the tower of states analysis (TOS) [1, 2]. The energy spectrum, i.e. the eigenvalues of
the Hamiltonian, of a finite system has a characteristic and systematic structure in a symmetry
broken phase: several eigenstates are quasi-degenerate on finite systems, become degenerate
in the thermodynamic limit and possess certain quantum numbers. The TOS analysis deals
with understanding the spectral structure of the Hamiltonian and predicting quantum numbers
of the groundstate manifold. Also on finite systems spontaneous symmetry breaking mani-
fests itself in the structure of the energy spectra which are accessible via numerical simulations.
Most prominently, the Exact Diagonalization method [3, 4] can exactly calculate these spectra,
including their quantum numbers, on moderate system sizes. Different well-established nu-
merical techniques like the Quantum Monte Carlo technique also allow for performing energy
level spectroscopy to a certain extent [5]. The predictions of TOS analysis are highly nontrivial
statements which can be used to unambiguously identify symmetry broken phases. Thus TOS
analysis is a powerful technique to investigate many condensed matter systems using numerical
simulations. The goal of these lecture notes is to explain the specific structure of energy spec-
tra and their quantum numbers in symmetry broken phases. The anticipated structure is then
compared to several actual numerical simulations using Exact Diagonalization.

These lecture notes have been written at the kind request of the organizers of the Julich 2016
Autumn School on Correlated Electrons [6]. The notes build on and complement previously
available lecture notes by Claire Lhuillier [2], by Gregoire Misguich and Philippe Sindzingre [7]
and by Karlo Penc and one of the authors [8].

The outline of these notes is as follows: in Section 2 we introduce the tower of states of con-
tinuous symmetry breaking and derive its scaling behaviour. We investigate a toy model which
shows most of the relevant features. Section 3 explains in detail how the multiplicities and quan-
tum numbers in the TOS can be predicted by elementary group theoretical methods. To apply
these methods we discuss several examples in Section 4 and compare them to actual numerical
data from Exact Diagonalization.

2
2 TOWER OF STATES

2 Tower of states
We start our discussion on spontaneous symmetry breaking by investigating the Heisenberg
model on the square lattice. Its Hamiltonian is given by
X
H=J Si Sj (1)
hi,ji

and is invariant under global SU(2) spin rotations, i.e. a rotation of every spin on each site with
the same rotational SU(2) matrix. Therefore the total spin
!2
X
S2tot = Si = Stot (Stot + 1)
i

is a conserved quantity of this model and every state in the spectrum of this Hamiltonian can
be labeled via its total spin quantum number Stot . The Heisenberg Hamiltonian on the square
lattice has the property of being bipartite: The lattice can be divided into two sublattices A and
B such that every term in Eq. (1) connects one site from sublattice A to sublattice B. It was
found out early [1] that the groundstate of this model bears resemblance with the classical Neel
state
|Neel class.i = | i

where the spin-ups live on the A sublattice and the spin-downs live on the B sublattice. The
total spin Stot is not a good quantum number for this state. From elementary spin algebra we
know that it is rather a superposition of several states with different total spin quantum numbers.
For example the 2-site state

|i |i |i + |i
|i = + = |Stot = 0, m = 0i + |Stot = 1, m = 0i
2 2

is the superposition of a singlet (Stot = 0) and a triplet (Stot = 1). Therefore if such a state were
to be a groundstate of Eq. (1) several states with different total spin would have to be degenerate.
It turns out that on finite bipartite lattices this is not the case: The total groundstate of the
Heisenberg model on bipartite lattices can be proven to be a singlet state with Stot = 0. This
result is known as Marshalls Theorem [911]. So how can the Neel state resemble the singlet
groundstate? To understand this we drastically simplify the Heisenberg model and investigate a
toy model whose spectrum can be fully understood analytically.

2.1 Toy model: the Lieb-Mattis model

By introducing the Fourier transformed spin operators
N
1 X ikxj
Sk = e Sj
N j=0

3
2.1 Toy model: the Lieb-Mattis model 2 TOWER OF STATES

we can rewrite the original Heisenberg Hamiltonian in terms of these operators as

X
H=J k Sk Sk (2)
kBZ

where k = cos(kx ) + cos(ky ) and the sum over k runs over the momenta within the first
Brillouin zone (BZ). Let k0 = (, ) be the ordering wavevector which is the dual to the
translations that leave the square Neel state invariant. We now want to look at the truncated
Hamiltonian
HLM = 2J S2(0,0) Sk0 Sk0


(3)

where we omit all Fourier components in Eq. (2) except k = (0, 0) and k0 = (, ). This model
is called the Lieb-Mattis model [10] and has a simple analytical solution. To see this we notice
that Eq. (3) can be written as
4J X
HLM = Si Sj (4)
N iA,jB

in real space, where A and B denote the two bipartite sublattices of the square lattice and each
spin is only coupled with spins in the other sublattice. The interaction strength is equal regard-
less of the distance between the two spins. Thus this model is not likely to be experimentally
relevant. Yet it will serve as an illustrative example how breaking the spin-rotational symmetry
manifests itself in the spectrum of a finite size system. We can rewrite Eq. (4) as
!
4J X X X
HLM = Si Sj Si Sj Si Sj
N i,jAB i,jA i,jB
4J 2
= (Stot S2A S2B )
N
This shows that the Lieb-Mattis model can be considered as the coupling of two large spins SA
and SB to a total spin Stot .
We find that the operators S2tot , Stot
z
, S2A and S2B commute with this Hamiltonian and therefore
the sublattice spins SA and SB as well as the total spin Stot and its z-component mtot are good
quantum numbers for this model. For a lattice with N sites (N even) the sublattice spins can be
chosen in the range SA,B {0, 1, . . . , N/4} and by coupling them

Stot {|SA SB |, |SA SB | + 1, . . . , SA + SB }

mtot {Stot , Stot + 1, . . . , Stot }

can be chosen1 . A state |Stot , m, SA , SB i is thus an eigenstate of the systems with energy

4J
E(Stot , m, SA , SB ) = [Stot (Stot + 1) SA (SA + 1) SB (SB + 1)] (5)
N
independent of m, so each state is at least (2Stot + 1)-fold degenerate.
1
This set of states spans the full Hilbertspace of the model.

4
2 TOWER OF STATES 2.1 Toy model: the Lieb-Mattis model

Tower of states We first want to consider only the lowest energy states for each Stot sector.
These states build the famous tower of states and collapse in the thermodynamic limit to a
highly degenerate groundstate manifold, as we will see in the following.
For a given total spin Stot the lowest energy states are built by maximizing the last two terms in
Eq. (5) with SA = SB = N/4 and

4J N
E0 (Stot ) = E(Stot , m, N/4, N/4) = Stot (Stot + 1) J( + 1) (6)
N 4
The groundstate of a finite system will thus be the singlet state with Stot = 0 2 . On a finite
system the groundstate is, therefore, totally symmetric under global spin rotations and does not
break the SU (2)-symmetry. In the thermodynamic limit N , however, the energy of all
these states scales to zero and all of them constitute to the groundstate manifold.
The classical Neel state with fully polarized spins on each sublattice can be built out of these
states by a linear combination of all the Stot levels with mtot = 0 [2]. All other Neel states
pointing in a different direction in spin-space can be equivalently built out of this groundstate
manifold by considering linear combinations with other mtot quantum numbers. In the thermo-
dynamic limit, any infinitesimal small field will force the Neel state to choose a direction and
the groundstate spontaneously breaks SU (2)-symmetry.
The states which constitute the groundstate manifold in the thermodynamic limit can be readily
identified on finite-size systems as well, where their energy and spin quantum number are given
by Eq. (6). These states are called the tower of states (TOS) or also Anderson tower, thin
spectrum and quasi-degenerate joint states [1, 1214].

Excitations The lowest excitations above the tower of states can be built by lowering the spin
of one sublattice SA or SB by one, see Eq. (5). Let us set SA = N/4 and SB = N/4 1 which
implies that Stot {1, 2, . . . , N/2 1}. We can directly compute the energy E1 (Stot ) of these
excited states for each allowed Stot . The energy gap to the tower of states

Eexc (Stot ) = E1 (Stot ) E0 (Stot ) = J

is constant3 . Hence, the lowest excitations of the Lieb-Mattis model are static spinflips. The
next lowest excitations are spinflips on both sublattices, SA = SB = N/4 1 with excitation
energy Eexc2 = 2J and Stot {0, 1, . . . , N/2 2}. We observe that only the energy gap of the
TOS levels vanishes in the thermodynamic limit, so the TOS indeed solely contributes to the
groundstate manifold.

Quantum Fluctuations When we introduced the Lieb-Mattis model Eq. (3) from the Heisen-
berg model Eq. (2) we neglected all Fourier components except of k = (0, 0) and k = k0 . This
2
The groundstate of the Heisenberg model Eq. (1) on a bipartite sublattice with equal sized sublattices is also
proven to be a singlet state Stot = 0 by Marshalls Theorem [11, 9, 10].
3
This is an artifact of the infinite-range interaction in the Lieb-Mattis model. In the original Heisenberg model
these modes become gapless magnon excitations.

5
 3 SYMMETRY ANALYSIS

was a quite crude approximation and it is not guaranteed that all results for the Lieb-Mattis
model will survive for the short-range Heisenberg model. To get some first results regarding
this question, we can introduce small quantum fluctuations on top of the Neel groundstate of
the Lieb-Mattis model and perform a perturbative spin-wave analysis in first order4 . This ap-
proach does not affect the scaling of the tower of states levels, but it has an important effect on
the excitations. They are not static particles anymore, but become spinwaves (magnons) with
a dispersion, which is linear around the ordering-wave vector k = k0 and k = (0, 0). On a
finite-size lattice the momentum space is discrete with a distance proportional to 1/L between
momentum space points, where L is the linear size of the system. The energy of the lowest
excitation above the TOS, the single magnon gap, therefore scales as Eexc J/L to zero5 .
As the scaling is, however, slower for d > 1-dimensional systems than the TOS scaling, these
levels do not influence the groundstate manifold in the thermodynamic limit. Furthermore, the
excitation of two magnons results in a two-particle continuum above the magnon mode.
The properties of the TOS and its excitations are summarized in Fig. 1. The left figure shows the
general properties of the finite-size energy spectrum which can be expected when a continuous
symmetry group is spontaneously broken in the thermodynamic limit. The right figure depicts
the energy spectrum for the Heisenberg model on a square lattice with N = 32 sites, obtained
with Exact Diagonalization. One can clearly identify the TOS, the magnon dispersion and the
many-particle continuum. The existence of a Neel TOS was not only confirmed numerically
for the Heisenberg model on the square lattice, but also with analytical techniques beyond the
simplification to the Lieb-Mattis model [1, 13, 14]. The different symbols in Fig. 1 represent
different quantum numbers related to the space-group symmetries on the lattice. In the next
section we will see that the structure of these quantum numbers depends on the exact shape of
the symmetry-broken state and we will learn how to compute them.

3 Symmetry analysis
In the analysis of excitation spectra from Exact Diagonalization on finite size simulation clusters
the TOS analysis is a powerful tool to detect spontaneous symmetry breaking. As we have
seen in the previous chapter explicitly for the Heisenberg antiferromagnet, symmetry breaking
implies degenerate groundstates in the thermodynamic limit. On finite size simulation clusters
this degeneracy is in general not an exact degeneracy. We rather expect a certain scaling of the
energy differences in the thermodynamic limit. We distinguish two cases:

Discrete symmetry breaking: In this case we have a degeneracy of finitely many states
in the thermodynamic limit. The groundstate splitting on finite size clusters scales as
exp(N/), where N is the number of sites in the system.
4
A more detailed discussion can be found in [2].
5
In the thermodynamic limit the single magnon mode is gapless and has linear dispersion around k = k0 and
k = (0, 0). It corresponds to the well-known Goldstone mode which is generated when a continuous symmetry is
spontaneously broken.

6
 What are the finite size manifestations of a continuous symmetry breaking ?

3 Low-energy
SYMMETRY dynamics of the order parameter
ANALYSIS
Theory: P.W. Anderson 1952, Numerical tool: Bernu, Lhuillier and others, 1992 -

Continuum
Energy

Magnons

Tower of
States

1/N 1/L

S(S+1)
Fig. 1: (Left): Schematic finite-size energy spectrum of an antiferromagnet breaking SU(2)
spin-rotational symmetry. The TOS levels are the lowest energy levels for each total spin S and
scale with 1/N to the groundstate energy. The low energy magnon excitations are seperated
from the TOS and a continuum of higher energy states and scale with 1/L. (Right): Energy
spectrum for the Heisenberg model on a square lattice from ED. The TOS levels are connected
by a dashed line. The single magnon dispersion (green boxes) with Stot {1, 2, . . . } are well
separated from the TOS and the higher multi-particle continuum. The different symbols repre-
sent quantum numbers related to space-group symmetries and agree with the expectations for a
Neel state (See section 3).

Continuous symmetry breaking: The groundstate in the thermodynamic limit is in-

finitely degenerate. The states belonging to this degenerate manifold collapse as 1/N
on finite size clusters as we have seen in section 2. It is important to understand that these
states are not the Goldstone modes of continuous symmetry breaking. Both the degen-
erate groundstate and the Goldstone modes appear as low energy levels on finite size
clusters but have different scaling behaviours.

The scaling of these low energy states can now be investigated on finite size clusters. More
importantly also their quantum numbers such as momentum, pointgroup representation or total
spin can be predicted [2, 7, 15]. The detection of correct scaling behaviour together with cor-
rectly predicted quantum numbers yields very strong evidence that the system spontaneously
breaks symmetry in the way that has been anticipated. This is the TOS method. In the fol-
lowing we will discuss how to predict the quantum numbers for discrete as well as continuous
symmetry breaking. The main mathematical tool we use is the character-formula from basic
group representation theory.
Lattice Hamiltonians like a Heisenberg model often have a discrete symmetry group arising
from translational invariance, pointgroup invariance or some discrete local symmetry, like a
spinflip symmetry. In this chapter we will first discuss the representation theory and the charac-
ters of the representations of space groups on finite lattices. We will then see how this helps us
to predict the representations of the degenerate ground states in discrete as well as continuous
symmetry breaking.

7
3.1 Representation theory for space groups 3 SYMMETRY ANALYSIS

3.1 Representation theory for space groups

For finite discrete groups such as the space group of a finite lattice the full set of irreducible
representations (irreps) can be worked out. Let us first discuss some basic groups. Lets con-
sider a n n square lattice with periodic boundary conditions and a translationally invariant
Hamiltonian like the Heisenberg model on it. In the following we will set the lattice spacing to
a = 1. The discrete symmetry group we consider is T = Zn Zn corresponding to the group
of translations on this lattice. This is an Abelian group of order n2 . Its representations can be
labeled by the momentum vectors k = ( 2i n
, 2j
n
), i, j {0, , n 1} which just correspond
to the reciprocal Bloch vectors defined on this lattice. Put differently, the vectors k are the re-
ciprocal lattice points of the lattice spanned by the simulation torus of our n n square lattice.
The character k of the k-representation is given by

k (t) = eikt

where t T is the vector of translation. This is just the usual Bloch factor for translationally
invariant systems.
Let us now consider a (symmorphic) space group of the form D = T PG as the discrete
symmetry group of the lattice where PG is the pointgroup of the lattice. For a model on a
n n square lattice this could for example be the dihedral group of order 8, D4 , consisting of
fourfold rotations together with reflections. The representation theory and the character tables
of these point groups are well-established. An example for such character tables can be found in
Tabs. 1 and 4 for the cyclic group C4 and the dihedral group D6 6 . Since D is now a product of
the translation and the point group we could think that the irreducible representations of D are
simply given by the product representations (k ) where k labels a momentum representation
and an irrep of PG. But here is a small yet important caveat. We have to be careful since
D is only a semidirect product of groups as translations and pointgroup symmetries do not
necessarily commute. This alters the representation theory for this product of groups and the
irreps of D are not just simply the products of irreps of T and PG. Instead the full set of irreps
for this group is given by (k k ) where k is an irrep of the so called little group Lk of k
defined as
Lk = {g PG; g(k) = k}

which is just the stabilizer of k in PG. For example, all pointgroup elements leave k = (0, 0)
invariant, thus the little group of k = (0, 0) is the full pointgroup PG. In general, this does
not hold for other momenta and only a subgroup of PG will be the little group of k. In Fig. 4
we show the k-points of a 6 6 triangular lattice together with its little groups as an example.
The K point in the Brillouin zone has a D3 little group, the M point a D2 little group. Having
discussed the represenation theory for (symmorphic) space groups we state that the characters

6
We follow the labeling scheme for point group representations according to Mulliken [16].

8
3 SYMMETRY ANALYSIS 3.2 Predicting irreducible representations

of these representations are simply given by

(k,k ) (t, p) = eikt k (p)

where t T , p PG and k denotes the character of the representation k of the little group
Lk .

3.2 Predicting irreducible representations in spontaneous symmetry break-

ing
Spontaneous symmetry breaking at T = 0 occurs when the groundstate |GS i of H in the
thermodynamic limit is not invariant under the full symmetry group G of H. We will call a
specific groundstate |GS i a prototypical state and the groundstate manifold is defined by
 i
VGS = span |GS i

i
where |GS i is the set of degenerate groundstates in the thermodynamic limit. This groundstate
manifold space can be finite or infinite dimensional depending on the situation. For breaking a
discrete finite symmetry, such as in the example given in section 4.1.2, this groundstate mani-
fold will be finite dimensional, for breaking continuous SO(3) spin rotational symmetry7 as in
section 4.2 it is infinite dimensional in the thermodynamic limit. For every symmetry g G
we denote by Og the symmetry operator acting on the Hilbert space. The groundstate manifold
becomes degenerate in the thermodynamic limit and we want to calculate the quantum numbers
of the groundstates in this manifold. Another way of saying this is that we want to compute the
irreducible representations of G to which the groundstates belong to. For this we look at the
action of the symmetry group G on VGS defined by
:G Aut(VGS ) (7)
i j


g 7 hGS |Og |GS i i,j
This is a representation of G on VGS , so every group element g G is mapped to an invertible
matrix on VGS . In general this representation is reducible and can be decomposed into a direct
sum of irreducible representations M
= n

These irreducible representations are the quantum numbers of the eigenstates in the ground-
state manifold and n are its respective multiplicities (or degeneracies). Therefore these irreps
constitute the TOS for spontaneous symmetry breaking [2]. To compute the multiplicities we
can use a central result from representation theory, the character formula
1 X
n = (g) Tr( (g)) (8)
|G| gG
7
The actual symmetry group of Heisenberg antiferromagnets is usually SU(2). For simplicity we only consider
the subgroup SO(3) in these notes which yields the same predictions for the case of sublattices with even number
of sites (corresponding to integer total sublattice spin).

9
 4 EXAMPLES

where (g) is the character of the representation and Tr( (g)) denotes the trace over the
representation matrix (g) as defined in Eq. (7). Often we have the case that

1 if O | 0 i = | i
0 g GS GS
hGS |Og |GS i =
0 otherwise

With this we can simplify Eq. (8) to what we call the character-stabilizer formula
1 X
n = (g) (9)
|Stab(|GS i)|
gStab(|GS i)

where
Stab(|GS i) {g G : Og |GS i = |GS i}

is the stabilizer of a prototypical state |GS i 8 . We see that for applying the character-stabilizer
formula in Eq. (9) only two ingredients are needed:

the stabilizer Stab(|GS i) of a prototypical state |GS i in the groundstate manifold

the characters of the irreducible representations of the symmetry group G

We want to remark that in the case of G = D C where D is a discrete symmetry group such
as the spacegroup of a lattice and C is a continuous symmetry group such as SO(3) rotations for
Heisenberg spins the Eqs. (8) and (9) include integrals over Lie groups additionally to the sum
over the elements of the discrete symmetry group D. Furthermore also the characters for Lie
groups like SO(3) are well-known. For an element R SO(3) the irreducible representations
are labeled by the spin S and its characters are given by
sin (S + 21 )
 
S (R) =
sin 2

where [0, 2] is the angle of rotation of the spin rotation R. We work out several exam-
ples for this case in section 4.2 and compare the results to actual numerical data from Exact
Diagonalization.

4 Examples
4.1 Discrete symmetry breaking
In this section we want to apply the formalism of section 3 to systems, where only a discrete
symmetry group is spontaneously broken but not a continuous one. In this case, the ground-
state of the system in the thermodynamic limit is described by a superposition of a finite number
8
In some cases, the orbit of the prototypical state G. |GS i = {g G : Og |GS i} does not span the full set
i
of degenerate groundstates |GS i. In this case, we have to find a set of prototypical states with different orbits,
such that the union of these orbits spans the full groundstate manifold. Then, Eq. (9) has to be applied to each
prototypical state, individually, and the final multiplicity is the sum of the individual results.

10
4 EXAMPLES 4.1 Discrete symmetry breaking

of degenerate eigenstates with different quantum numbers. On finite-size systems, however, the
symmetry cannot be broken spontaneously and a unique groundstate will be found. The other
states constituting to the degenerate eigenspace in the thermodynamic limit exhibit a finite-size
energy gap which is exponentially small in the system size N , eN/ . The quantum
numbers of these quasi-degenerate set of eigenstates are defined by the symmetry-broken state
in the thermodynamic limit.

4.1.1 Introduction to valence-bond solids

In section 2 we have seen that the classically ordered Neel state is a candidate to describe the
groundstate of the antiferromagnetic Heisenberg model Eq. (1) with J > 0 in the thermody-
namic limit on a bipartite lattice. The energy expectation value of this state on a single bond is
eNeel = J/4.
The state which minimizes the energy of a single bond is, however, a singlet state |S = 0i
formed by the two spins on the bond with energy eVB = 3J/4, called a valence bond (VB) or
dimer. A valence bond covering of an N -site lattice can then be described by a tensor product
of N/2 VBs, where each site belongs to exactly one VB9 . Another possible candidate for the
thermodynamic groundstate of Eq. (1) is then a superposition of all possible VB coverings with
only nearest neighbour VBs. Such states do not break the SU (2) spin-rotational symmetry as
Stot = 0 and are in general not eigenstates of the Hamiltonian: Acting with the operator Si Sj
between sites i and j belonging to two different VBs changes the VB configuration.
This manifold of VB coverings is highly degenerate. As the VB coverings are in general not
eigenstates of the Hamiltonian, they encounter quantum fluctuations. The energy corrections
due to these fluctuations are typically not equivalent for different coverings, although the bare
energies are identical. The VB coverings with the largest energy gain are selected by the fluc-
tuations as the true groundstate configurations. If this order-by-disorder mechanism [19, 20]
selects regular patterns of VB coverings, the discrete lattice symmetries can be spontaneously
broken in the thermodynamic limit, and a valence bond solid (VBS) can be formed. Fig. 2
and Fig. 3 show two different VBS states on the square lattice. VBSs show no long-range spin
order, but long-range dimer-correlations h(Sa Sa0 )(Sb Sb0 )i where a, a0 and b, b0 label sites
on individual dimers. In section 4.1.2 we will see how different VBS states can be identified
and distinguished by the quantum numbers of the quasi-degenerate groundstate manifold on
finite-size systems.
The groundstate of the Heisenberg model Eq. (1) on the square lattice is not a VBS but a Neel
state, which has a lower variational energy already on the classical level. Nevertheless, several
models featuring VBS groundstates are known in 1- and 2-D [2125]. Interestingly, in [26]
a model was proposed, which shows a direct continuous quantum phase transition between a
Neel state and a VBS. This transition exhibits very exotic, non-classical behaviour and is called
deconfined quantum critical point [27].
9
The set of all possible valence bond coverings with arbitrary length spans the full Stot = 0 sector of the models
Hilbert space and is overcomplete [17, 18].

11
4.1 Discrete symmetry breaking 4 EXAMPLES

4.1.2 Identification of VBSs from finite-size spectra

Columnar valence-bond solid A columnar VBS (cVBS) on a square lattice is shown in

Fig. 2. Four equivalent states can be found, indicating that there will be a four-fold quasi-
degenerate groundstate manifold. A cVBS breaks the translational and point-group symmetries
of an isotropic SU(2)-invariant Hamiltonian on the lattice spontaneously but not the continuous
spin symmetry group since it is a singlet and thus invariant under spin rotations.

Fig. 2: The four columnar VBS coverings of a square lattice. Valence bonds (spin singlets) are
indicated by blue ellipses.

In the following we use Eq. (9) to compute the symmetry sectors of the groundstate manifold.
The discrete symmetry group we consider is

G = D = T PG

where T = Z2 Z2 = {1, tx , ty , tx ty } are the non-trivial lattice translations with translation

vectors
t1 = (0, 0), tx = (1, 0), ty = (0, 1), txy = (1, 1)
and PG = C4 denotes the point-group of four-fold lattice rotations10 . To compute the ground-
state symmetry sectors we do not need to consider the full symmetry group G but only the
stabilizer Stab(|cV BS i), leaving one of the states in Fig. 2 unchanged. Without loss of gener-
ality we choose the first covering as prototype |cV BS i. The stabilizer is given by

Stab(|cV BS i) = {1 1} {1 C2 } {ty 1} {ty C2 }

where C2 denotes the rotation about an angle around the center of a plaquette.
The irreducible representations (irreps) of the group of lattice translations T can be labelled by
the allowed momenta k

k Irreps(T ) = {(0, 0), (, 0), (0, ), (, )},

and the corresponding characters for an element t T are

k (t) = eikt .

The irreps (called A, B and E, see [16]) and characters for the point-group C4 are tabulated in
Tab. 1.
10
The dihedral group D4 is also a symmetry group of the model. For the sake of simplicity we decided to only
consider the subgroup C4 in this section.

12
4 EXAMPLES 4.1 Discrete symmetry breaking

C4 1 C4 C2 (C4 )3
A +1 +1 +1 +1
B +1 -1 +1 -1
Ea +1 +i -1 -i
Eb +1 -i -1 +i

Table 1: Character table for pointgroup C4 .

Using the character-stabilizer formula Eq. (9) we can now reduce the representation induced
by the state |cV BS i to irreducible representations to get the quantum numbers of the quasi-
degenerate groundstate manifold. Let us explicitely consider k = (0, 0) as an example:
1 X
n(0,0)A = n(0,0)B = A (d)k=(0,0) (d)
|Stab(|cV BS i)|
dStab(|cV BS i)
1
1 eik(0,0) + 1 eik(0,0) + 1 eik(0,1) + 1 eik(0,1) = 1

=
4
1 X
n(0,0)Ea = n(0,0)Eb = B (d)k=(0,0) (d)
|Stab(|cV BS i)|
dStab(|cV BS i)
1
1 eik(0,0) + (1) eik(0,0) + 1 eik(0,1) + (1) eik(0,1) = 0

=
4
Eventually, the cVBS covering will be described by a four-fold quasi-degenerate groundstate
manifold with the following quantum numbers11 .

(|cV BS i) = (0, 0)A (0, 0)B (, 0)A (0, )A.

VBS states are a superposition of spin singlets on the lattice, therefore the spin quantum number
for all levels in the groundstate manifold must be trivial, Stot = 0.

Staggered valence-bond solid The columnar VBS is not the only regular dimer covering of
the square lattice. Another possible regular covering is the staggered VBS (sVBS), where again
four equivalent configurations span the groundstate manifold. One of these configurations is
shown in Fig. 3.
Similarly, also the sVBS spontaneously breaks the translational and point-group symmetries
of an isotropic Hamiltonian, but not the spin-rotational symmetry. Following the same steps
as before we can compute the quantum numbers of the four quasi-degenerate groundstates for
the sVBS. The stabilizer turns out to be different to the case of the cVBS and thus also the
decomposition into irreps yields a different result:

(|sV BS i) = (0, 0)A (0, 0)B (, )Ea (, )Eb .

11
The little group k for the momenta k = (0, ) and k = (, 0) is only the subgroup C2 of 2-fold roations
of the symmetry group C4 considered in this example. The irrep called A therefore denotes the trivial irreducible
representation of C2 for these momenta. When computing the multiplicities for these momenta one should also
note, that two prototype states have to be considered to span the full groundstate manifold under the symmetry
elements of C2 .

13
4.2 Continuous symmetry breaking 4 EXAMPLES

Fig. 3: One of the four identical staggered VBS coverings on the square lattice.

Tab. 2 shows a comparison of the irreducible representations in the groundstate manifold of the
cVBS and sVBS states.

Irreps cVBS sVBS

(0, 0) A 1 1
(0, 0) B 1 1
(, 0) A 1 0
(0, ) A 1 0
(, )Ea 0 1
(, )Eb 0 1

Table 2: Multiplicities of the irreducible representations in the four-fold degenerate groundstate

manifolds of the columnar and staggered VBS on a square lattice.

By a careful analysis of the quasi-degenerate states and their quantum numbers on finite systems
it is thus possible to identify and distinguish different VBS phases which spontaneously break
the space group symmetries in the thermodynamic limit.

4.2 Continuous symmetry breaking

In this section we give several examples of systems breaking continuous SO(3) symmetry. We
discuss the introductory example of the square lattice Heisenberg antiferromagnet, calculate the
irreps in the TOS and compare this to actual energy spectra from Exact Diagonalization on a
finite lattice in section 4.2.1. In section 4.2.2 we discuss three magnetic orders on the triangular
lattice and an extended Heisenberg model where all of these are stabilized. We present results
from Exact Diagonalization and compare the representations in these spectra to the predictions
from TOS analysis. Finally, we introduce quadrupolar order and show that also this kind of
symmetry breaking can be analyzed using the TOS technique in section 4.2.3.

4.2.1 Heisenberg antiferromagnet on the square lattice

We now give a first example how the TOS method can be applied to predict the structure of the
tower of states for magnetically ordered phases. We look at the Neel state of the antiferromagnet
on the bipartite square lattice with sublattices A and B. A prototypical state in the groundstate

14
4 EXAMPLES 4.2 Continuous symmetry breaking

manifold is given by
|i = | i

where all spins point up on sublattice A and down on sublattice B. The symmetry group G =
D C of the model we consider is a product between discrete translational symmetry D =
Z2 Z2 = {1, tx , ty , txy } and spin rotational symmetry C = SO(3). We remark that we restrict
our translational symmetry group to D = Z2 Z2 instead of D0 = Z Z because the Neel
state transforms trivially under two-site translations (tx )2 , (ty )2 . Thus, only the representations
of D0 trivial under two-site translations are relevant; these are exactly the representations of D.
Put differently we only have to consider the translations in the unitcell of the magnetic structure
which in the present case can be chosen as a 2-by-2 cell. Furthermore, we will for now neglect
pointgroup symmetries like rotations and reflections of the lattice to simplify calculations. At
the end of this section we give results where also these symmetry elements are incorporated.
The groundstate manifold VGS we consider are the states related to |i by an element of the
symmetry group G, i.e.
VGS = {Og |i ; g G}

The symmetry elements in G that leave our prototypical state |i invariant are given by two sets
of elements:

No translation in real space or a diagonal txy translation together with a spin rotation
Rz () around the z-axis with an arbitrary angle .

Translation by one site, tx or ty , followed by a rotation Ra () of 180 around an axis

a z perpendicular to the z-axis.

So the stabilizer of our prototype state |i is given by

Stab(|i) = {1 Rz ()} {txy Rz ()} {tx Ra ()} {ty Ra ()}

The representations of the discrete symmetry group can be labeled by four momenta k
{(0, 0), (0, ), (, 0), (, )} with corresponding characters

k (t) = eikt

where t denotes the translation vector corresponding to t. The continuous symmetry group
we consider is the Lie group SO(3). Its representations are labeled by the total spin S. The
character of the spin-S representation is given by

sin (S + 21 )
 
S (R) =
sin 2

where [0, 2] is the angle of rotation of the element R SO(3). We see that spin rotations
with different axes but same rotational angle give rise to the same character. The representations

15
4.2 Continuous symmetry breaking 4 EXAMPLES

of the total symmetry group G = D C are now just the product representations of D and C.
Therefore also the characters of representations of G are the product of characters of D and C.
We label these representations by (k, S) where k denotes the lattice momentum and S the total
spin. To derive the multiplicities of the representations (k, S) in the groundstate manifold, we
now apply the character-stabilizer formula, Eq. (9). In the case of the square antiferromagnet
this yields

Z2 Z2
1 1
n(k,S) = eik0 dS (Rz ()) + eik(ex +ey ) dS (Rz ())
4 |Rz ()| 4 |Rz ()|
0 0
Z2 2
1 1
Z
+ eikex daS (Ra ()) + eikey daS (Ra ())
4 |Ra ()| 4 |Ra ()|
0 0

We compute
Z2
|Rz ()| = |Ra ()| = d = 2
0

Z2 Z2 Z2 S
sin (S + 21 )
 
1 1 1 X
dS (Rz ()) = d = d eil = 1 (10)
2 2 sin 2 2 l=S
0 0 0

and
Z2 Z2
sin (S + 12 )
 
1 1
daS (Ra ()) = da = (1)S (11)
2 2 sin 2
0 0

Putting this together gives the final result for the multiplicities of the representations in the tower
of states
(
1 1 if S even
1 1 + 1 1 + 1 (1)S + 1 (1)S =


n((0,0),S) =
4 0 if S odd
(
1 0 if S even
1 1 + 1 1 1 (1)S 1 (1)S =


n((,),S) =
4 1 if S odd
1
1 1 1 1 + 1 (1)S 1 (1)S = 0


n((0,),S) =
4
1
1 1 1 1 1 (1)S + 1 (1)S = 0


n((,0),S) =
4
Tab. 3 lists the computed multiplicities of the irreducible representations where additionally the
D4 point group was considered in the symmetry analysis. These irreps and their multiplicities
exactly agree with the irreps and multiplicities in the TOS of the square lattice Heisenberg
model from ED in Fig. 1. The spectroscopic predictions together with the numerical data thus
constitute a firm and solid evidence of Neel order.

16
4 EXAMPLES 4.2 Continuous symmetry breaking

S .A1 M .A1
0 1 0
1 0 1
2 1 0
3 0 1

Table 3: Multiplicities of irreducible representations in the TOS for the Neel Antiferromagnet
on a square lattice.

4.2.2 Magnetic order on the triangular lattice

On the triangular lattice several magnetic orders can be stabilized. The Heisenberg nearest
neighbour model has been shown to have a 120 Neel ordered groundstate where spins on
neighbouring sites align in an angle of 120 [28,29]. Upon adding further second nearest neigh-
bour interactions J2 to the Heisenberg nearest neighbour model with interaction strength J1 it
was shown that the groundstate exhibits stripy order for J2 /J1 & 0.18 [30]. Here spins are
aligned ferromagnetically along one direction of the triangular lattice and antiferromagnetically
along the other two. Interestingly, it was shown that a phase exists between these two magnetic
orders whose exact nature is unclear until today. Several articles propose that in this region
an exotic quantum spin liquid is stabilized [3134]. In a recent proposal two of the authors
established an approximate phase diagram of an extended Heisenberg model with further scalar
chirality interactions J Si (Sj Sk ) [35] on elementary triangles. The Hamiltonian of this
model is given by
X X X
H = J1 Si Sj + J2 Si Sj + J Si (Sj Sk ) (12)
hi,ji hhi,jii i,j,k4

Amongst the already known 120 Neel and stripy phases an exotic Chiral Spin Liquid and a
magnetic tetrahedrally ordered phase were found. Here we will only discuss the magnetic
orders appearing in this model. The non-coplanar tetrahedral order has a four-site unitcell where
four spins align such that they span a regular tetrahedron. In this chapter we discuss the tower
of states for the three magnetic phases in this model.
First of all Fig. 4 shows the simulation cluster used for the Exact Diagonalization calculations
in [35]. We chose a N = 36 = 6 6 sample with periodic boundary conditions. This sample
allows to resolve the momenta , K and M , amongst several others in the Brillouin zone.
The K and M momenta are the ordering vectors for the 120 , stripy and tetrahedral order.
Furthermore, this sample features full sixfold rotational as well as reflection symmetries (the
latter only in the absence of the chiral term, i.e. J = 0). Its pointgroup is therefore given by
the dihedral group of order 12, D6 . The little groups of the individual k vectors are also shown
in Fig. 4. For our tower of states analysis we now want to consider the discrete symmetry group

D = T D6

where T is the translational group of the magnetic unitcell. The full set of irreducible represen-
tations of this symmetry group is given by the set (k k ) where k denotes the momentum and

17
4.2 Continuous symmetry breaking 4 EXAMPLES

Fig. 4: (Left): Simulation cluster for the Exact Diagonalization calculations. (Center): Brillouin
zone of the triangular lattice with the momenta which can be resolved with this choice of the
simulation cluster. Different symbols denote the little groups of the corresponding momentum.
(Right): TOS for the 120 Neel order on the triangular lattice. The symmetry sectors and
multiplicities fulfill the predictions from the symmetry analysis (See Tab. 5). One should note,
that the multiplicities grow with Stot for non-collinear states.

k is an irrep of the little group associated to k. The points , K and M give rise to the little
groups D6 , D3 and D2 (the dihedral groups of order 12, 8, and 4), respectively. For the stripy
and tetrahedral order we can choose a 2 2 magnetic unitcell, and a 3 3 unitcell for the 120
Neel order. The spin rotational symmetry gives rise to the continuous symmetry group

C = SO(3)

We can therefore label the full set of irreps as (k, k , S) where S denotes the total spin S
representation of SO(3). Similarly to the previous chapter we now want to apply the character-
stabilizer formula, Eq. (9), to determine the multiplicities of the representations forming the
tower of states. The characters of the irreps (k, k , S) are given by

sin (S + 21 )
 
ikt
(k,k ,S) (t, p, R) = e k (p)
sin 2

where again [0, 2] is the angle of rotation of the spin rotation R. The characters of

D6 1 2C6 2C3 C2 3d 3v
A1 1 1 1 1 1 1
A2 1 1 1 1 -1 -1
B1 1 -1 1 -1 1 -1
B2 1 -1 1 -1 -1 1
E1 2 1 -1 -2 0 0
E2 2 -1 -1 2 0 0

Table 4: Character table for pointgroup D6 .

18
4 EXAMPLES 4.2 Continuous symmetry breaking

Fig. 5: (Left): TOS for the stripy phase on the triangular lattice. The multiplicities for each
even/odd Stot are constant for collinear phases. (Right): TOS for the tetrahedral order on the
triangular lattice.

the pointgroup D6 are given in Tab. 4. We skip the exact calculations which follow closely
the calculations performed in the previous chapter, although now pointgroup symmetries are
additionally taken into account. The results are summarized in Tab. 5. We remark that the

120 Neel stripy order tetrahedral order

S .A1 .B1 K.A1 .A1 .E2 M.A .A .E2a .E2b M.A
0 1 0 0 1 1 0 1 0 0 0
1 0 1 1 0 0 1 0 0 0 1
2 1 0 2 1 1 0 0 1 1 1
3 1 2 2 0 0 1 1 0 0 2

Table 5: Multiplicities of irreducible representations in the Anderson tower of states for the
three magnetic orders on the triangular lattice defined in the main text.

tetrahedral order is stabilized only for J 6= 0 where the model in Eq. (12) does not have
reflection symmetry any more since the term Si (Sj Sk ) does not preserve this symmetry.
Therefore we used only the pointgroup C6 of sixfold rotation in the calculations of the tower of
states for this phase.
If we compare these results to Figs. 4,5 we see that these are exactly the representations appear-
ing in the TOS from Exact Diagonalization for certain parameter values J2 and J . This is a
strong evidence that indeed SO(3) symmetry is broken in these models in a way described by
the 120 Neel, stripy and tetrahedral magnetic prototype states.
It is worth noting, that the sum of the multiplicities is constant with Stot for collinear phases,
e.g. the stripy order shown here, whereas it is increasing for non-collinear orders.

4.2.3 Quadrupolar order

All examples of continuous symmetry breaking we have discussed so far spontaneously broke
SO(3) symmetry but exhibited a magnetic moment. In the following we will show examples

19
4.2 Continuous symmetry breaking 4 EXAMPLES

of phases that do not exhibit any magnetic moment but break spin-rotational symmetry anyway
and discuss the influences on the tower of states. We will restrict our discussion to quadrupolar
phases in S = 1 models here, a broader introduction to nematic and multipolar phases can be
found in [8].

Quadrupolar states We denote the basis states for a single spin S = 1 with S z = 1, 1, 0
as |1i , |1i , |0i. In contrast to the usual S = 1/2 case not each basis state can be obtained by
a SU(2) rotation of any other basis state. The state |0i, for example cannot be obtained by a
rotation of |1i or |1i as it has no orientation in spin-space at all, h0|S |0i = 0 [8]. The state |0i
can, however, be described as a spin fluctuating in the x y plane in spin space as

h0|(S x )2 |0i = h0|(S y )2 |0i = 1, h0|(S z )2 |0i = 0

We can thus assign a director d along the z-axis to this state. SU(2) rotations change the
director of such a state, but not its property of being non-magnetic. These states are identified
as quadrupolar states as they can be detected by utilizing the quadrupolar operator [8]
2
Q = S S + S S S(S + 1)
3
To study the possible formation of an ordered quadrupolar phase on a lattice, where the direc-
tors of the quadrupoles on each lattice site follow a regular pattern, we consider the bilinear-
biquadratic model with Hamiltonian
X
H= J Si Sj + Q (Si Sj )2 (13)
hi,ji

and S = 1. The second term in Eq. (13) can be rewritten in terms of the elements of Q which
can be rearranged into a 5-component vector Q such that
4
Qi Qj = 2(Si Sj )2 + Si Sj (14)
3
The expectation value of Eq. (14) for quadrupolar states on sites i and j is given in terms of
their directors di,j [8]:
2
hQi Qj i = 2 (di dj )2
3
Therefore, the second term in Eq. (13) favours regular patterns of the directors of quadrupoles.
When such states are formed, they spontaneously break SU(2) symmetry without exhibiting
any kind of magnetic moment. The first term in Eq. (13), on the other hand, favours magnetic
spin ordering as we have already discussed in previous sections.
The phase diagram of Eq. (13) on the triangular lattice shows extended ferromagnetic, antifer-
romagnetic (120 ), ferroquadrupolar (FQ) and antiferroquadrupolar (AFQ) ordered phases. In
the FQ phase quadrupoles on each lattice site are formed with all directors pointing in a single
direction, whereas the directors form a 120 structure in the AFQ phase. In the following, we
will show that the FQ and AFQ phases can be identified and distinguished from the spin ordered
phases using the TOS analysis on finite clusters.

20
4 EXAMPLES 4.2 Continuous symmetry breaking

TOS for quadrupolar phases The TOS for the FQ and AFQ phases can be expected to show
similar behaviour as the TOS for magnetically ordered states as both spontaneously break the
spin-rotational symmetry. If we identify the symmetry-broken quadrupolar phases with their
directors pointing in any direction in spin-space we can perform the symmetry analysis of the
TOS levels in a very similar manner as for spin-ordered systems in the previous sections. There
is, however, one important thing to consider: The directors should not be considered to be
described with vectors, but with axes; a quadrupole is recovered (up to a phase) by rotations
about an angle around any axis a in the xy-plane:
a
eiS |0i = |0i (15)
Thus, the stabilizer in Eq. (9) is different for quadrupolar phases and the TOS shows a different
structure. This property makes it possible to distinguish, e.g., a magnetic 120 phase from its
quadrupolar counterpart, the AFQ phase using TOS analysis.
A prototype |i for the FQ phase is a product states of quadrupoles with directors in z-direction.
This state does not break any space-group symmetries, so only the trivial irreps of the space
group, k = = (0, 0), A1, will be present in the TOS. The remaining stabilizer of the spin-
rotation group is a rotation around the z-axis about an arbitrary angle and a rotation about an
angle around any axis lying in the xy-plane,
Stab(| i) = {Rz (), Ra ()}
The multiplicities in the TOS can then be computed as
 Z 2 Z 2 
1 1 N 1
nS = dS (Rz ()) + (1) daS (Ra ())
2 |Rz ()| 0 |Ra ()| 0
1
1 + (1)N (1)S


=
2
where the integrals have already been computed in Eqs. (10) and (11). The system size depen-
dent factor (1)N is imposed from Eq. (15). To sum up, the TOS for the FQ phase has single
levels for even (odd) S with trivial space-group irreps and no levels for odd (even) S sectors
when N is even (odd)12 . The absence of odd (even) S levels is caused by the invariance of
quadrupoles under -rotation and distinguishes the TOS for a FQ phase from a usual ferromag-
netic phase. In Fig. 6 the computed TOS for the model Eq. (13) in the FQ phase is shown on the
left. It shows the expected quantum numbers and multiplicities in the TOS and also an easily
identifiable magnon branch below the continuum.
The symmetry analysis for the AFQ phase can be performed in a similar manner and shows
a similar structure to the magnetic 120 -Neel phase, but again levels are deleted for the AFQ.
In this case, however, not all odd levels are deleted but some levels in both, odd and even, S
sectors. Tab. 6 shows the multiplicities of irreps in the TOS of the AFQ model in comparison to
the magnetic 120 -Neel state for even N . Fig. 6 shows the simulated TOS for the AFQ phase
for the bilinear-biquadratic model Eq. (13). The symmetry sectors and multiplicities agree with
the predictions.
12
For the simple case of the FQ phase one can also easily calculate the decomposition of a state |S = 1, m = 0i
|S = 1, m = 0i . . . into states |Stot , m = 0i with the use of Clebsch-Gordan coefficients.

21
 5 OUTLOOK

Fig. 6: Tower of states for the ferroquadrupolar (left) and antiferroquadrupolar (right) states on
a triangular lattice with N = 12 sites from Exact Diagonalization. The single-magnon branch
for the FQ phase is highlighted with green boxes.

AFQ 120 Neel

S .A1 .B1 K.A1 .A1 .B1 K.A1
0 1 0 0 1 0 0
1 0 0 0 0 1 1
2 0 0 1 1 0 2
3 0 1 0 1 2 2

Table 6: Irreducible representations and multiplicities for the AFQ phase compared to the mag-
netic 120 -Neel phase.

5 Outlook
In the previous sections we have discussed prominent features of the energy spectrum for states
which spontaneously break the spin-rotational symmetry in the thermodynamic limit. We have
seen that on finite-size systems the energy spectra of such states exhibit a tower of states (TOS)
structure. The tower of states scales as Stot (Stot + 1)/N and generates the groundstate manifold
in the thermodynamic limit N , which is indispensible to spontaneously break a sym-
metry. The quantum numbers of the levels in the TOS depend on the particular state which is
formed after the symmetry breaking and can be predicted using representation theory.

As a generalization to the SU(2)-symmetric Heisenberg model, Eq. (1), one can introduce SU(n)
Heisenberg models with n > 2. Such models can experimentally be realized by ultracold mul-
ticomponent fermions in an optical lattices. When the on-site repulsion is strong enough, the
Hamiltonian can be effectively described by an SU(n) symmetric permutation model on the
lattice [36]. If the exchange couplings are antiferromagnetic, SU(n) generalized versions of the
Neel state might be realized as groundstates, which then spontaneously break the SU(n) sym-
metry of the Hamiltonian. On finite systems this becomes again manifest in the emergence of
a tower of states, where the scaling is found to be proportional to C2 (n)/N [3740, 36]; C2 (n)

22
5 OUTLOOK

denotes the quadratic Casimir operator of SU(n)13 . The symmetry analysis of the levels in the
TOS can in principle be performed similar to the case of SO(3) discussed in these notes but the
symmetry group and its characters have to be replaced with the more complicated group SU(n).

On the other side, it can be also interesting to study models where the continuous symme-
try group is smaller. In real magnetic materials, the isotropic Heisenberg interaction is often
accompanied by other interactions which, when they are strong enough, might reduce the sym-
metry group of spin rotations to O(2); only spin rotations around an axis are a symmetry of the
system and can be spontaneously broken in the thermodynamic limit. This symmetry group is
also interesting in the field of ultracold gases, as BECs spontaneously break an O(2) symmetry
by choosing a phase. Tower of states can also be found in this case and the quantum numbers
and multiplicities of the TOS levels can be computed in a similar fashion [15].

= i - Square = 12 + 3
2 i - Triangular
16

14

12
N /c

10

(E E0)

6
Z2 even, ED/QMC
4
Z2 odd, ED/QMC
2 Z2 even, -exp
Z2 odd, -exp
0
0 1 2 2 5 0 1 3 2

Fig. 7: Universal torus spectrum for a continuous quantum phase transition in the 3D Ising uni-
versality class. Full symbols denote numerical results while empty symbols denote -expansion
results. The dashed line shows a dispersion with the speed of light.

We have seen, that the energy spectrum of Hamiltonians on finite lattices may contain a lot of
information about the system. One can identify groundstates which will spontaneously break
discrete as well as continuous symmetries in the thermodynamic limit and by imposing a classi-
cal state as symmetry broken state one can even predict the quantum numbers and multiplicities
of the levels in the tower of states or in the quasi-degenerate groundstate manifold. When we
impose an additional interaction to a system with spontaneously broken groundstate, e.g. a mag-
netic field, it is possible that a continuous quantum phase transition (cQPT) from the ordered
state to a disordered state appears for some critical ratio of the couplings. Such cQPTs are inter-
esting as they can be described by universal features which do not depend on most microscopic
details of the model. Interestingly, the energy spectrum on finite systems can even be used to

identify and characterize cQPTs. It is given by universal numbers times 1/L, where L = N
13
For n = 2 the quadratic Casimir operator C2 = Stot (Stot + 1).

23
 5 OUTLOOK

is the linear size of the lattice. The quantum numbers of the energy levels show universal fea-
tures and are qualitatively related to the operator content of the underlying critical field theory,
although the relation between them is not yet fully understand for non-flat geometries, like a
torus [41, 42]. The critical spectrum for the transverse field Ising model on a torus is shown in
Fig. 7. It is a fingerprint for the 3D Ising cQPT.

24
REFERENCES REFERENCES

References
[1] P. W. Anderson, Phys. Rev. 86, 694 (1952). doi:10.1103/PhysRev.86.694
http://link.aps.org/doi/10.1103/PhysRev.86.694

[2] C. Lhuillier, arXiv:cond-mat pp. 161190 (2005). doi:10.1007/3-540-45649-X 6

http://arxiv.org/abs/cond-mat/0502464

[3] A. M. Lauchli: Numerical Simulations of Frustrated Systems (Springer Berlin Heidelberg,

Berlin, Heidelberg, 2011), pp. 481511. doi:10.1007/978-3-642-10589-0 18
http://link.springer.com/10.1007/978-3-642-10589-0_18

[4] A. W. Sandvik, A. Avella, and F. Mancini: In AIP Conf. Proc. (2010), Vol. 1297, pp. 135
338. doi:10.1063/1.3518900
http://arxiv.org/abs/1101.3281

[5] H. Suwa and S. Todo, Phys. Rev. Lett. 115, 080601 (2015). doi:10.1103/PhysRevLett.
115.080601
http://link.aps.org/doi/10.1103/PhysRevLett.115.080601

[6] E. Pavarini, E. Koch, J. van den Brink, and G. Sawatzky: Quantum Materials: Experi-
ments and Theory, Modeling and Simulation, Vol. 6 (Forschungszentrum Julich, Julich,
2016)
http://juser.fz-juelich.de/record/819465

[7] G. Misguich and P. Sindzingre, J. Phys. Condens. Matter 19, 145202 (2007). doi:10.1088/
0953-8984/19/14/145202
http://iopscience.iop.org/0953-8984/19/14/145202

[8] K. Penc and A. M. Lauchli: Spin Nematic Phases in Quantum Spin Systems
(Springer Berlin Heidelberg, Berlin, Heidelberg, 2011), pp. 331362. doi:10.1007/
978-3-642-10589-0 13
http://link.springer.com/10.1007/978-3-642-10589-0_13

[9] W. Marshall, Proc. R. Soc. A Math. Phys. Eng. Sci. 232, 48 (1955). doi:
10.1098/rspa.1955.0200
http://rspa.royalsocietypublishing.org/cgi/doi/10.1098/
rspa.1955.0200

[10] E. Lieb and D. Mattis, J. Math. Phys. 3, 749 (1962)

http://scitation.aip.org/content/aip/journal/jmp/3/4/10.
1063/1.1724276

[11] A. Auerbach: Interacting Electrons and Quantum Magnetism. Graduate Texts in

Contemporary Physics (Springer New York, New York, NY, 1994). doi:10.1007/

25
REFERENCES REFERENCES

978-1-4612-0869-3
http://link.springer.com/10.1007/978-1-4612-0869-3

[12] T. A. Kaplan, W. von der Linden, and P. Horsch, Phys. Rev. B 42, 4663 (1990). doi:
10.1103/PhysRevB.42.4663
http://link.aps.org/doi/10.1103/PhysRevB.42.4663

[13] P. Hasenfratz and F. Niedermayer, Zeitschrift fr Phys. B Condens. Matter 92, 91 (1993).
doi:10.1007/BF01309171
http://link.springer.com/10.1007/BF01309171

[14] P. Azaria, B. Delamotte, and D. Mouhanna, Phys. Rev. Lett. 70, 2483 (1993). doi:10.1103/
PhysRevLett.70.2483
http://link.aps.org/doi/10.1103/PhysRevLett.70.2483

[15] I. Rousochatzakis, A. M. Lauchli, and F. Mila, Phys. Rev. B 77, 094420 (2008). doi:
10.1103/PhysRevB.77.094420
http://link.aps.org/doi/10.1103/PhysRevB.77.094420

[16] R. S. Mulliken, J. Chem.Phys. 23, 1997 (1955). doi:http://dx.doi.org/10.1063/1.1740655

[17] S. Liang, B. Doucot, and P. W. Anderson, Phys. Rev. Lett. 61, 365 (1988). doi:10.1103/
PhysRevLett.61.365
http://link.aps.org/doi/10.1103/PhysRevLett.61.365

[18] C. Lhuillier and G. Misguich: In C. Lacroix, P. Mendels, and F. Mila (Eds.) Introd. to
Frustrated Magn. (Springer Berlin Heidelberg, Berlin, Heidelberg, 2011), Springer Series
in Solid-State Sciences, Vol. 164, pp. 2341. doi:10.1007/978-3-642-10589-0
http://link.springer.com/10.1007/978-3-642-10589-0_2

[19] E. F. Shender, Sov. Phys. JETP 56, 178 (1982)

[20] C. L. Henley, Phys. Rev. Lett. 62, 2056 (1989). doi:10.1103/PhysRevLett.62.2056

http://link.aps.org/doi/10.1103/PhysRevLett.62.2056

[21] J. Fouet, P. Sindzingre, and C. Lhuillier, Eur. Phys. J. B 20, 241 (2001). doi:10.1007/
s100510170273
http://link.springer.com/10.1007/s100510170273

[22] A. Lauchli, S. Wessel, and M. Sigrist, Phys. Rev. B 66, 014401 (2002). doi:10.1103/
PhysRevB.66.014401
http://link.aps.org/doi/10.1103/PhysRevB.66.014401

[23] A. Lauchli, J. C. Domenge, C. Lhuillier, P. Sindzingre, and M. Troyer, Phys. Rev. Lett. 95,
137206 (2005). doi:10.1103/PhysRevLett.95.137206
http://link.aps.org/doi/10.1103/PhysRevLett.95.137206

26
REFERENCES REFERENCES

[24] M. Mambrini, A. Lauchli, D. Poilblanc, and F. Mila, Phys. Rev. B 74, 144422 (2006).
doi:10.1103/PhysRevB.74.144422
http://link.aps.org/doi/10.1103/PhysRevB.74.144422

[25] A. Gelle, A. M. Lauchli, B. Kumar, and F. Mila, Phys. Rev. B 77, 014419 (2008). doi:
10.1103/PhysRevB.77.014419
http://link.aps.org/doi/10.1103/PhysRevB.77.014419

[26] A. W. Sandvik, Phys. Rev. Lett. 98, 227202 (2007). doi:10.1103/PhysRevLett.98.227202

http://link.aps.org/doi/10.1103/PhysRevLett.98.227202

[27] T. Senthil, A. Vishwanath, L. Balents, S. Sachdev, and M. P. A. Fisher, Science (80-. ).

303, 1490 (2004). doi:10.1126/science.1091806
http://www.sciencemag.org/cgi/doi/10.1126/science.1091806

[28] T. Jolicoeur, E. Dagotto, E. Gagliano, and S. Bacci, Phys. Rev. B 42, 4800 (1990). doi:
10.1103/PhysRevB.42.4800
http://link.aps.org/doi/10.1103/PhysRevB.42.4800

[29] A. V. Chubukov and T. Jolicoeur, Phys. Rev. B 46, 11137 (1992). doi:10.1103/PhysRevB.
46.11137
http://link.aps.org/doi/10.1103/PhysRevB.46.11137

[30] P. Lecheminant, B. Bernu, C. Lhuillier, and L. Pierre, Phys. Rev. B 52, 6647 (1995). doi:
10.1103/PhysRevB.52.6647
http://link.aps.org/doi/10.1103/PhysRevB.52.6647

[31] Y. Iqbal, W.-J. Hu, R. Thomale, D. Poilblanc, and F. Becca, Phys. Rev. B 93, 144411
(2016). doi:10.1103/PhysRevB.93.144411
http://link.aps.org/doi/10.1103/PhysRevB.93.144411

[32] R. Kaneko, S. Morita, and M. Imada, J. Phys. Soc. Jpn. 83, 093707 (2014). doi:10.7566/
JPSJ.83.093707
http://journals.jps.jp/doi/10.7566/JPSJ.83.093707

[33] W.-J. Hu, S.-S. Gong, W. Zhu, and D. N. Sheng, Phys. Rev. B 92, 140403 (2015). doi:
10.1103/PhysRevB.92.140403
http://link.aps.org/doi/10.1103/PhysRevB.92.140403

[34] Z. Zhu and S. R. White, Phys. Rev. B 92, 041105 (2015). doi:10.1103/PhysRevB.92.
041105
http://link.aps.org/doi/10.1103/PhysRevB.92.041105

[35] A. Wietek and A. M. Lauchli, Phys. Rev. B 95, 035141 (2017). doi:10.1103/PhysRevB.
95.035141
http://link.aps.org/doi/10.1103/PhysRevB.95.035141

27
REFERENCES REFERENCES

[36] P. Nataf and F. Mila, Phys. Rev. Lett. 113, 127204 (2014). doi:10.1103/PhysRevLett.113.
127204
http://link.aps.org/doi/10.1103/PhysRevLett.113.127204

[37] K. Penc, M. Mambrini, P. Fazekas, and F. Mila, Phys. Rev. B 68, 012408 (2003). doi:
10.1103/PhysRevB.68.012408
http://link.aps.org/doi/10.1103/PhysRevB.68.012408

[38] T. A. Toth, A. M. Lauchli, F. Mila, and K. Penc, Phys. Rev. Lett. 105, 265301 (2010).
doi:10.1103/PhysRevLett.105.265301
http://link.aps.org/doi/10.1103/PhysRevLett.105.265301

[39] P. Corboz, A. M. Lauchli, K. Penc, M. Troyer, and F. Mila, Phys. Rev. Lett. 107, 215301
(2011). doi:10.1103/PhysRevLett.107.215301
http://link.aps.org/doi/10.1103/PhysRevLett.107.215301

[40] P. Corboz, M. Lajko, K. Penc, F. Mila, and A. M. Lauchli, Phys. Rev. B 87, 195113 (2013).
doi:10.1103/PhysRevB.87.195113
http://link.aps.org/doi/10.1103/PhysRevB.87.195113

[41] M. Schuler, S. Whitsitt, L.-P. Henry, S. Sachdev, and A. M. Lauchli, Phys. Rev. Lett. 117,
210401 (2016). doi:10.1103/PhysRevLett.117.210401
http://link.aps.org/doi/10.1103/PhysRevLett.117.210401

[42] S. Whitsitt and S. Sachdev, Phys. Rev. B 94, 085134 (2016). doi:10.1103/PhysRevB.94.
085134
http://link.aps.org/doi/10.1103/PhysRevB.94.085134

28